Strengthening defenses against AI fraud

– Kriti Bansal, vice president finance and accounting, AlphaPoint

Ask an Expert

Q. Can advisors work with AI to ensure clients are safe against fraud?

A.Yes, but AI should support advisors and not act as autonomous decision-maker. It can flag unusual wallet behavior, suspicious contracts, phishing patterns and risky approvals before damage happens. The biggest vulnerability today is granting AI agents direct, unmitigated wallet permission and this can turn the agent itself into a massive attack vector for social engineering or bad on-chain data.

Q. What security in real time looks like in the age of AI?

A. In the age of AI the real time security needs to be predictive and proactive and not reactive. Real-time security means warnings before signing, continuous wallet monitoring, instant alerts on abnormal activity and blocking risky approvals before funds can move.

Q. How can a money manager automate a defense layer that acts as a continuous threat monitor?

A.Money managers must move away from legacy externally owned wallets and transition to programmable smart accounts such as ERC-4337 or EIP-7702. This transition allows one to write automated, programmatic security guardrails directly at the account level. They can use automated monitoring for wallets, approvals, contract risks, transaction patterns and exposure limits, with human escalation for anything unusual.